大数据驱动的高端装备故障诊断与安全服役保障的理论与技术

Because of the great amount of the high-end equipment groups monitored and the large number of the sensors for each equipment, massive data is acquired by the high sampling frequency after the long-time service of equipment. Such massive data pushes the field of fault diagnosis to enter the era of big data, which brings new challenges in the theories and methods of fault diagnosis. For example, the data is big but fragmentary, the fault feature extraction relies on much prior knowledge and diagnostics expertise, the diagnosis accuracy of the shallow diagnosis model is low, and the anti-interference ability of individual remaining useful life prediction methods is weak. According to the guide of NSFC-Zhejiang integration of informatization and industrialization union funds, this project rises to these challenges to meet the industrial requirement of safe servcie of high-end equipment driven by big data. The study contents contain the acquisition of complete big data, the intelligent representation of fault information, the deep identification of fault modes, and the model-data joint prediction of remaining useful life, which aim to explore the real-time mapping relationship between the monitoring big data and the fault development of high-end equipment, reveal the principle of the intelligent analysis and characterization of big data, and address the mechanism of deep identification of multiple fault modes and safe service evaluation. The research results of this project will provide the basic theories and the key techniques for the fault diagnosis and safe service of high-end equipment in the era of big data, and benefit the companies using high-end equipment in Zhejiang as well as the whole country, further promoting deep integration of informatization and industrialization.

由于高端装备群规模大、每台装备测点多、数据采样频率高、装备服役历时长，所以获取了海量的监测数据，驱动故障诊断领域进入了“大数据”时代。面对海量数据，已有机械故障诊断理论与方法遇到了新挑战：数据大而不全呈“碎片化”、诊断特征提取受制于人为经验、浅层模型故障识别准确率低、单一寿命预测方法抗干扰能力弱等。本项目瞄准上述挑战，依据NSFC-浙江两化融合联合基金指南，以大数据驱动的高端装备安全服役工程需求为导向，研究监测大数据的完备获取、故障信息的智能表征、故障模式的深度识别、剩余寿命的数模联动预测等关键内容，旨在探明高端装备故障演化过程与监测大数据的实时映射关系，揭示大数据蕴含信息的智能解析与表征原理，阐述多故障模式的深度识别与安全服役评估机制，形成大数据下的高端装备智能故障诊断与安全服役的基础理论与关键技术，服务于浙江乃至全国相关行业的高端装备，进一步促进工业化与信息化的深度融合。

本项目围绕大数据下高端装备故障诊断与安全服役面临的关键科学问题与技术难点，研究了监测大数据完备获取、故障信息解析与智能表征、故障智能诊断与预测等基础理论与关键技术。主要研究成果与创新点包括：（1）提出了高精度模型数值仿真与装备实测数据相结合的监测大数据完备获取方法，解决了监测大数据“大而不全”的难题，探明了高端装备故障演变过程与大数据的实时映射关系。（2）提出了监测大数据质量保障与信息表征技术，实现了大数据提纯与复杂故障特征的自适应提取，揭示了大数据故障信息的智能解析与自适应表征原理。（3）构建了大数据驱动的深度智能诊断模型，实现了多标记故障的智能识别，阐述了装备多故障模式的深度识别机制。（4）建立了高端装备的迁移智能诊断模型，克服了大数据下故障标签数据稀缺的智能诊断应用难题。（5）构建了数模联动的智能寿命预测模型，实现了装备剩余寿命的准确预测，建立了数据退化信息与模型经验知识之间的交互匹配模式。（6）与浙江企业建立了“机械装备健康监测联合实验室”，并向全球学者公开发布了高端轴承加速寿命数据（国内首个、国际第三个），累计下载过万次。（7）与浙江多家企业开展校企合作，研发了大数据智能诊断与预测系统，并在浙江乃至全国的风电能源、高端制造、交通运输等领域推广应用。（8）出版中文专著1部；共发表有本项目资助号标注的期刊论文73篇、会议论文13篇；完成博士后出站报告1篇、博士学位论文7篇、硕士学位论文15篇；申请国际发明专利2项、授权国家发明专利28项，其中技术转让1项；参与制定故障诊断国家标准4项；项目负责人雷亚国教授获国家技术发明二等奖（第2）、陕西省自然科学一等奖（第1）、中国青年科技奖、教育部青年科学奖（全国每年不超过10人）等奖励、国家杰青基金资助，连续入选科睿唯安全球高被引科学家、爱思唯尔中国高被引学者。合作单位负责人向家伟教授获中国核能行业协会科技进步三等奖（第3）。